EP0608945B1 - Star sensor with CCD-array, method of detection and application to reposition a spacecraft - Google Patents

Description

The present invention relates to a star finder for analysis of the image of at least one luminous object, comprising a sensor of position which implements an element detector matrix charge transfer photosensitive (DTC) and image memory in which is transferred, after each integration time Ti and during a reading time T1 in the form of digital samples, the image detected by said matrix and where it replaces the previous image.

The invention also relates to a method of detection by accumulation of thumbnails defined by variable position windows, on the DTC matrix, and it advantageously applies to stellar registration of a maneuvering satellite.

In order to obtain the two information (in coordinates Cartesian) to completely determine the direction of the object luminous observed, in particular a star, it is known, for example from U.S. Patent No. 4,430,673 to use a detector photosensitive consisting of a charge transfer detection matrix. This detector is presented as a mosaic of image elements (pixels) divided into two parts: a first half, the photosensitive area is designed to receive and detect light radiation from the object in question, and the second half, covered with an opaque film, serves as analog memory in which are stored, by charge transfer, the information received by the first half of the matrix during a duration Ti known as integration time. the two halves of the matrix detector preferably have separate charge transfer controls. A read register then receives, always by charge transfer, line by line, the information in the memory area, and then these information is provided in analog form to an output stage at from which they can be read pixel by pixel so that they can reconstruct the image previously obtained on the photosensitive area under in the form of digital samples, stored in an image memory. For stellar aiming in particular, a first mode of operation of the device, called search mode, consists in defining, in a sub-matrix the so-called search window contained in the matrix described more top, the position, to the nearest pixel, of the luminous object sought after, being understood that the search window has been previously defined from a sensor inertial, for example, installed on the same device as the sighting of the luminous object (generally on a spacecraft). Any either the magnitude, the luminous object observed, in general a star, results, on the detection matrix at a light point whose diameter is the order of a few pixels assuming a slight defocus, performed by the lens. In search mode, the pixel which thus receives all the signal of the emissive source (the star) sought achieves the indicated goal above giving the direction of the star by its coordinates with a accuracy equal to 0.5 times the side of a pixel.

A DTC matrix star finder is designed to cover a square or rectangular field of a few degrees over a few degrees, a degree itself being covered by about 100 pixels on the matrix.

To ensure attitude measurement regardless of orientation of the line of sight of the viewfinder in space, it is necessary that, for any orientation taken at random, at least one measurable star is present in the field viewfinder. Experience shows that, to do this, it is necessary to introduce into the catalog of viewfinder stars, stars of medium magnitude, typically 5 to 6, the catalog of stars required, on the 4 steradian II of the celestial sphere, being of the order of 1000 stars.

If the angular movement of the viewfinder relative to the stars is slow, which is the case for example when the viewfinder is mounted on a non-maneuvering satellite (which travels naturally in its orbit), the integration on the DTC matrix of a star, even a dim light, does not no problem because the drag effect is weak and the signal to S / N noise remains sufficient, with an integration time Ti of the order of one second, so that the energy barycenter of the signal detected in each integration window can be identified as the star trace sought, which belongs to the aforementioned star catalog. Such a state of the art that star sights are used in combination with inertial sensors for precise attitude measurements of artificial satellites is described for example in US Pat. No. 4,617,634.

Thus, with existing viewfinders, the measurement frequency is limited by the intensity of the star to be measured. We are then led to low frequency measuring devices, at most a few hertz. This therefore does not make it possible to detect the position of star (s) as soon as the angular movements of the viewfinder support platform become order one degree per second and more. The aim, in particular the phase of detection, then becomes impossible, the drag effect being too marked, in other words, the available energy being spread over too many pixels. Angular movements of the order of a degree per second and most meet on maneuvering satellites, satellites observation in particular.

A known solution to solve the technical problem indicated in the previous paragraph is to use as stellar sights image dissecting tubes. Such tubes allow you to aim for stars of greater magnitude, of the order of 7 or 8, and of angular movement faster than DTC viewfinders: a small measurement module a star tracking filter on the tube allows a servo-control of the position of the axis of the tube on the direction of the star. However, the image dissecting tube has the defects inherent in a analog technology: complexity, fragility, weight, large dimensions, so many disadvantages, moreover, increased in a spatial context.

According to US-A patent 4,672,220 which relates to the same technical field, only one light integration is carried out on the DTC sensor, followed by a single reading of all the pixels in the matrix DTC; at the signal level of each pixel is substituted, in the direction of lines, this same level increased by the signal sum of several pixels which are adjacent to it, which amounts to carrying out an integration, in the sense mathematical, according to a sliding window of small width.

In patent application EP-A 0 499 320, the problem posed is different from that which is at the start of the present invention. he it is a question of identifying a luminous task, without trail effect, which extends over several pixels due to a slight defocus deliberately created by the optical system. Again, the technical solution found requires a only reading of the DTC matrix.

An object of the invention is to detect weakly stars luminous with rapid angular movements by means of a DTC matrix viewfinder.

Another purpose is to allow an (aero) spacecraft to readjust its position on dimly lit stars at using DTC array finders mounted on platforms directly attached to said spacecraft.

According to the invention, these aims are achieved by virtue of the fact that the star finder indicated in the first paragraph is remarkable in that, for the detection and analysis of the image of dimly luminous objects in rapid movement, said finder stars having, at any time, available, as for the displacement of said luminous object relative to said position sensor, assistance information provided by reference frame reference means linked to said luminous object, it further comprises a memory section digital window dedicated to the storage of a window, containing the image of said light object, of predetermined dimensions and of variable position on said detector matrix, selection means for, during each integration time Ti _{n + 1} , select in the image contained in memory, said readjusted window relating to the previous integration time Ti _n , adding means for adding, pixel sample by ec pixel swivel, said window readjusted to the window previously contained in said digital memory section and to fill the latter with the result of addition thus obtained, before the end of the integration of the image integrated at time Ti _{n + 1} , the total number of windows accumulated, making it possible to obtain a signal-to-noise ratio satisfactory for the image of the target star, being equal to P, and that it is provided with calculation means for determining the position of the image of the luminous object on said matrix, by calculation of barycenters and vector addition of these barycenters.

The invention is primarily based on the idea that, the number of stars of medium magnitude being sufficient to achieve the goals sought, there is no need to use stars that are too large magnitude that would be undetectable by existing DTC matrices, whatever their mode of operation. Furthermore, the invention is made possible by the fact that the search window, sized on the precision in abscissa and ordinate of the position of the star on the matrix, and its displacement, are very small compared to the latter, which allows the accumulation of several tens of windows without leaving the matrix detection. The star is pursued, on the matrix, at a frame frequency adapted to the range of motion. The signal can then be integrated by addition of thumbnails, one thumbnail per window, over a period of time (period) long enough, on the order of a second, which achieves a trade-off between the desired precision, ie the S / N ratio, and the amplitude of the movement of the star on the matrix. The star signal collected at each frame (each integration) being a priori insufficient to allow a satisfactory detection, the pursuit, which comes down to a selection of successive windows, requires knowing the apparent movement of the star on the detector; this information can be obtained from inertial sensors, consisting either of an inertial unit external to the viewfinder, which continuously provides attitude information directly to the sight, either by one (or two) middle class inertial sensor (s), specific and integrated into said star viewfinder position sensor, which constitutes two possible variants for means of locating repository.

• pendant le temps Ti_n d'intégration de l'image d'ordre n le prétraitement, par l'unité centrale, des informations d'assistance, consistant à traduire dans le référentiel du viseur lesdites informations sur le mouvement relatif détecté entre le viseur et l'image qu'il reçoit sur sa matrice DTC,
• pendant le temps Tt_n suivant, court par rapport à Ti_n, le transfert en mémoire analogique de la matrice de l'image d'ordre n de la matrice intégrée pendant le temps Ti_n,
• puis le calcul d'assistance, pour la détermination de la position de la fenêtre qui contient l'imagette d'ordre n,
• puis la lecture de l'image d'ordre n pendant le temps Tl_n qui se termine avant la fin du temps Ti_n+1,
• pendant un temps Ta_n, compris dans la durée Tl_n, l'acquisition de l'imagette d'ordre n par des moyens de sélection actionnés à partir des résultats du calcul d'assistance, et
• la sommation de l'imagette d'ordre n aux imagettes précédentes dont la somme est contenue dans une section de mémoire de fenêtre, et le remplacement dans cette section de l'ancienne imagette somme, d'ordre n - 1, par la nouvelle, d'ordre n, ainsi que la mémorisation dans ladite section de la position de la fenêtre d'ordre n.

A method for analyzing the image of at least one star, using a star finder with a matrix of DTC elements as defined above, comprising an image memory, into which is transferred, after each integration time Ti, during a reading time Tl, in the form of digital samples, the image detected by said matrix, also comprising a central unit intended for calculations and a sequencer, is remarkable in that, for the detection of 'a dimly luminous and fast-moving star, an accumulation of P thumbnails in as many windows of predetermined dimensions readjusted, on said DTC matrix, on said star, is carried out, which comprises the following steps:

• during the time Ti _n of integration of the order image n the preprocessing, by the central processing unit, of assistance information, consisting in translating into the viewfinder repository said information on the relative movement detected between the viewfinder and the image it receives on its DTC matrix,
• during the next time Tt _n , short with respect to Ti _n , the transfer into analog memory of the matrix of the image of order n of the integrated matrix during the time Ti _n ,
• then the assistance calculation, for determining the position of the window which contains the thumbnail of order n,
• then the reading of the image of order n during the time Tl _n which ends before the end of the time Ti _{n + 1} ,
• during a time Ta _n , included in the duration Tl _n , the acquisition of the thumbnail of order n by selection means actuated from the results of the assistance calculation, and
• the summation of the thumbnail of order n to the preceding thumbnails whose sum is contained in a section of window memory, and the replacement in this section of the old thumbnail sum, of order n - 1, by the new, of order n, as well as the memorization in said section of the position of the window of order n.

A thumbnail accumulation process that follows precedent is remarkable in that said central unit determines the average position of said star on the matrix of DTC elements by vector addition of the geometric barycenter vector of the P positions of window on said matrix and the coincident energy barycenter vector with the position of the light object analyzed in the thumbnail image of order P.

The following description with reference to the accompanying drawings, all given by way of nonlimiting example, will make it clear how the invention can be realized.

Figure 1 is the block diagram of a star finder according to the invention.

Figure 2 is a diagram explaining the addition process of thumbnails made by the star viewfinder of figure 1.

Figure 3 shows, in A, B and C a thumbnail reconstructed according to the invention, at the rate of a single thumbnail at A, of the cumulation 4 thumbnails in B, and a total of 16 thumbnails in C.

Figure 4 is an explanatory time diagram of the workflow for processing digital samples representative of the pixels of the DTC matrix of the viewfinder.

FIG. 5 represents an optical copying device of a mechanical repository in a stellar viewfinder according to the invention.

In Figure 1 is shown schematically a viewfinder of stars proper, referenced 1. For its known part this viewfinder includes a lens (not shown) which forms a luminous object, more particularly of the star E an image E ', covering approximately 4 pixels on a photosensitive detector 3. Detector 3 is a matrix of detectors Photosensitive to Charge Transfer (pixels DTC) placed in substance at focus of the lens. This detector is divided into two parts: a first half, the photosensitive zone 4 is intended to receive and detect by integration of photons, the light radiation coming from the object targeted. The second half 5 serves as analog memory in which are stored, by charge transfer, for a short time Tt before Ti, the information received by the first half 4 during an integration time Ti. The two halves 4 and 5 of the matrix detector 3 have commands for separate charge transfer from a sequencer 6. A register of reading 7 then receives, always by line-by-line load transfer under the sequencer 6 command, the information contained in the zone memory and then this information is provided in analog form after amplification through an amplifier 8, to a video chain ensuring the digital conversion of the signal from which they can be read pixel by pixel, for a time Tl, so as to be able to reconstruct the image previously obtained on the photosensitive zone 4, in the form digital samples. In a classic stellar viewfinder, the image digital is supplied directly to a process image 13. The elements already mentioned electronic 6, 9 and 13 indicated above belong to a electronic viewfinder assembly providing all the focusing functions signal and interface. This electronics, which can be grouped together in a housing integral or not with the viewfinder 1 proper comprises in in addition to a central unit 12 and the associated memories (RAM / PROM) 13 (which serve among other things as image memories), to manage the whole and perform numerical processing (calculations). Sequencer 6 generates all the clock signals necessary for the operation of the detector 3 but also to video channel 9 and to the preprocessing of Support information explained below. Power supplies and easements of the whole are symbolized by a block 10 (cooling of the matrix to DTC in particular).

A common bus 14 allows communication bidirectional between elements 12, 13 and 6. To finish with the classic part of the viewfinder, a bus qualified as: user dialogue bus, 15, is connected to bus 14.

According to the invention, the viewfinder further comprises, in its part electronics, selection means and addition means, grouped together in a specific unit 16 as well as, associated with the image memory 13, a window memory section itself doubles 17, 170. The block of selection-addition 16 is controlled by the sequencer 6. It receives the digital samples from video channel 9, and communicates bidirectionally with memories 13, 17, 170 and with bus 14. In addition, the viewfinder receives on bus 14, coming from a benchmark locating unit 18, inertial assistance information on a bus 19. The unit 18 can be external or internal to the viewfinder, which is symbolized by the broken line 20. We suppose at first that it is an external unit, for example a control system attitude, such as an isolated-core inertial unit, in which case the assistance information consists of attitude information under the form of increments of angular movement. This is the form of assistance the simplest, but the viewfinder is no longer autonomous and it appears mechanical coupling problems such as a rigid connection necessary between the standards of the stellar viewfinder and of the inertial sensor, and of the constraints information relating to the information flow and the transfer times of data.

The process of detecting a low emissive source brightness (magnitude 5 to 6) and fast movement (from one to several degrees per second) is described below with reference to Figure 2. In the case of stellar aiming, it rests on the fact that the stars constitute a frame of reference which can be considered as inertial on the time scale of missions concerned (typically a few years).

The projection of the image of the star in the area photosensitive 4, due to the movement, produces a light streak 21 which, in the case considered, can extend over several tens of pixels per second. The width of this streak, given by the diameter of the spot image, is of the order of two pixels to allow barycentric calculation, even in the absence of movement, which represents the point of classic operation of a star finder. The treatment of assistance information from inertial sensors and provided by the bus 19 (figure 1) allows, with a constant drift, to follow the trace of the star on the photosensitive zone 4. This treatment essentially consists in a projection in the plane of zone 4 of the apparent movement of the star. In practice, it is the platform supporting the viewfinder that is animated of movement relative to the sky. The purpose of the measurement is to determine this drift which represents the projection, on zone 4, of the misalignment between the known position (attitude) of the platform from benchmark locating means (inertial sensors) and its attitude real in the benchmark that constitutes the sky, this last attitude being measured, according to the invention, by adding thumbnails which constitute under successive images of the images formed successively on the area photosensitive 4. In fact, the assistance thus established makes it possible to move a measurement window 22 which blindly follows the position of the star during each integration in zone 4. The selection of each window is performed by unit 16 (figure 1). Indeed, the dimensions of the window, typically 30 square pixels are such that they correspond to the maximum error that can occur on the position of the reconstructed star on zone 4 from assistance information, this theoretical position being in the center of each window such as 22. Given the nature of the detector which is a photosensitive matrix, the displacement is limited in fineness to the nearest pixel and is only possible with each frame (with each new signal integration). To be more precise, the dimensions of window 22 are determined by the uncertainty about the position of the star, the displacement of this star for the duration of an integration, the diameter of the spot image of the star and the margins necessary to apply the algorithms calculation (first order by adding the corresponding dimensions). The integration time Ti of each frame (elementary integration) is adapted to the amplitude of the movement during this same period so as to that the energy distribution obtained is optimal, that is to say that the measurement noise is minimal. This optimum corresponds to a displacement 23 (exaggerated in Figure 2), slightly greater than the diameter of the spot image, typically 2 to 3 pixels. The thumbnails thus obtained, 25, 26, 27, 28, i.e. the portions of images delimited by the measurement window 22, are superimposed and added successively to each other, as symbolically represented in 30. At each integration, the addition partial is carried out by unit 16 (Figure 1) and stored in the window storage section 17. At the same time, the coordinates of the windows (preferably the abscissa and zero-ordinate pixel of the window) is also kept in memory 170, associated with memory 17. In variant, these coordinates can be immediately added to prepare the following calculations. The addition of the thumbnails is continued until obtaining a sufficient S / N ratio to ensure the detection of the position true of the star inside the window. The representation made at the top of Figure 2 is purposely inaccurate, for educational purposes. In the practical, the number of windows along this trail is significantly more high, typically 4 to 12 times larger (16 to 50 cumulative images) and the photosensitive zone 4 has dimensions approximately tenfold of those represented (hundreds of pixels on the abscissa and on the ordinate).

The magnitude of the difference between the expected position and the position real of the star (drift) is represented, in 30, by the segment OO ', the point O being the center of the window and the point O 'the energy barycenter of the signal distribution resulting from the addition of the thumbnails. If the amplitude of the difference to be measured is too large, it may be necessary to set performs a preliminary search mode to detect the rough position the star image in the window (search for the maximum signal pixel for example). After the search, the precise position of the star image on the photosensitive zone 4 is obtained by linearly combining: the abovementioned barycenter, O ', preferably identified with respect to the original pixel P of the window, that is the vector PO ', defined by its Cartesian coordinates, and the average of the positions of the windows, that is the geometric barycenter Q, on zone 4, window positions, identified with respect to the original pixel R of zone 4, that is the vector RQ, defined by its Cartesian coordinates. In the particular case where the elementary integration times Ti are not not equal, the geometric barycenter is performed by weighting the positions windows by the corresponding integration times. Calculations barycentric, performed by the central unit 12 (FIG. 1), are symbolized, in FIG. 2, by block 31, for the energy barycenter, and by block 32, for the geometric barycenter. In block 33 is obtained the position average of the star in the sensor frame of reference (zone 4) over the entire duration of the measurement (sum of P elementary integrations), by addition abscissas, ordinates respectively, vectors RQ and PO '.

This result, compared to the integration of the measured movement by the inertial sensors over the same period of time, allows to readjust, in a manner known per se, the reference frame of the platform with respect to the stellar vault.

The result of the accumulation of the thumbnails in a 13 × 13 pixel window and for a unit of amplitude arbitrary: in A, it is the first thumbnail stored in the memory section 17. The star signal is not discernible from noise, this which shows that conventional star detection would be impossible. In B, where 4 thumbnails have been added, a significant trace appears in the northwest quadrant of the window. This trace can be that of the star sought but with an average probability, insufficient, because the noise level could easily reach the maximum amplitudes obtained, of the order of 10. In C, which represents the sum of 16 thumbnails, it is confirmed that this is indeed the star signal in the northwest quarter of the window, the maximum amplitude, for the pixel: x = 6, y = 11, being equal to 46. In practice, the accumulation of thumbnails is continued until either obtained an S / N ratio equal to 4 or 5 times the standard deviation. We can thus be leads to values of P of the order of 50.

FIG. 4 illustrates the typical time sequence of the main operations characteristic of the operating mode of the stellar viewfinder according to the invention. The elementary integrations (integration of order n, n + 1, at times Ti _n , Ti _{n + 1,} ...) and the corresponding readings of the matrix are interleaved, the reading of the image n, during time Tl _n (line B), being carried out during the following integration, of order n + 1. This reading mode is made possible by the architecture of the DTC matrices, such as for example the matrix marketed by the company Thomson-CSF under TH 7863, which are in the form of a mosaic of picture elements divided into two parts, as described above in the third paragraph. These two parts can be geographically distinct (frame transfer matrix, as shown in Figure 1) or nested (interline transfer matrix). At least one read register receives, still by charge transfer and line by line, the information stored in the analog memory area, which is then directed to an output stage from which it can be read, pixel by pixel so as to ability to reconstruct the image previously obtained on the photosensitive area. This arrangement makes it possible to integrate the signal with a minimum loss of time, that is to say a loss reduced at the time Tt of transferring the image from the first half to the second half of the matrix, this time Tt, due to the interline reading cycles being of the order of 100 times less than the integration time Ti. Reading of the image, during the time T1, is carried out, as indicated above, while the new integration of the signal is in progress. Simultaneously with the elementary integration, the inertial assistance information represented by arrows on the line EE, are sampled, at a frequency approximately tenfold of the frequency of the elementary integrations, and pretreated, according to their nature (this is generally increments of angular positions). This preprocessing, which takes place (line F) during a time Tp _n included in the elementary integration time of the same order number (Ti _n ) consists, in the electronic part of the viewfinder of FIG. 1, in converting the inertial assistance data received into data usable on the detection matrix. For that, it is necessary to carry out an average of the information of assistance during time Ti _n of the changes of coordinate system (of coordinate systems) and the corrections of angular values in a mathematical frame of reference, in this case in Cartesian coordinates on the matrix, it being understood that the initial position of the star, at the start of the registration operation is known and provided in the form of particular assistance information by the inertial sensors, for example via the bus 15 (FIG. 1). The assistance information is then in a form which can be used to establish the position of the measurement window. This assistance calculation, which notably includes pixel rounding calculations, and which is carried out during the time Tas _n (line G) at the start of the integration time following Ti _{n + 1} , must be carried out before the phase of reading of the image (before Tl _n ) to allow the pixels of interest to be selected, that is to say the acquisition of the window, at time Ta _n , included in the duration Tl _n (line VS). Finally (line D) this window of pixels is added to the current cumulative stored in the memory section 17 (Figures 1 and 2). It is advantageous, at this stage, to carry out the addition in real time to avoid the storage of too much information. Indeed, the typical dimensions of a window are a thousand pixels and the number of thumbnails a few tens, all leading to several tens of thousands of information (signal and pixel coordinates).

So far, only external inertial assistance has been considered. According to another preferred embodiment, the means for locating repository 18 (figure 1) consist of specific inertial sensors integrated into the stellar viewfinder. It is thus possible to obtain assistance internal which makes it possible to get rid of the rigid connection which is, if not, necessary between the stellar sight and the inertial sights. The result however, increased complexity of the stellar viewfinder. take in account the relatively low precision and drift requirements to ensure viewfinder assistance, these sensors can be middle class, such as miniature gyrometer (a two-axis gyroscope or two single-axis gyros axis), or tuning fork gyrometer. It is also possible to use two coupled accelerometers, arranged at the two ends of the axis of the sight.

A third embodiment, which can be deduced in particular respective methods of securing the inertial unit and the viewfinder on the spacecraft allows the implementation of the invention. In this case, the assistance information is obtained directly from the matrix under shape of sights, by optical copying of a mechanical reference frame the attitude is known, in the measurement reference frame. It is by example of the heart of a cardan inertial unit for example, like mechanical repository, connected to a platform via a mechanical suspension, while the viewfinder is fixedly mounted on this platform.

In FIG. 5, the mechanical reference frame 35 and the viewfinder 1 are supposed to have weak relative movements between them limited by example to relative vibration movements resulting from a suspension. The standard 35 includes a test projector (not shown), test targets which are injected into the field of view of viewfinder 1 by means of a optical invariant 36 (of the type: cube corner for example) along an axis optic 37. The optic invariant 36 may itself have a suspension independent of that of elements 1 and 35 and it includes a semi-transparent blade 38 which ensures the mixing between the signal of star E and test patterns representative of the mechanical reference system 35. In this particular case assistance, the viewfinder 1 itself ensures the constitution of information of attitude he needs by measuring the positions of the sights on the area photosensitive 4. For example, the measurement window can be defined as having a predetermined standard position, with respect to two sights, the sights remaining outside the window to prevent them from coming confuse with the star sought. This realization requires an adaptation, within the reach of the skilled person, the preprocessing of information assistance, aiming to allow, during each frame, the measurement of points corresponding targets. In addition, the intensity of these sights must be adapted so that the measurement accuracy is sufficient for each frame. The displacement observed between each integration is comparable to motion increments from previous forms of inertial assistance.

We can use this latter embodiment when it is not possible to ensure a rigid connection between the repository of the stellar viewfinder and that of inertial sensors.

More generally, optical copying and inertial measurements can be used in addition in the case of a repository non-inertial mechanics whose proper attitude is determined by inertial sensors. In this case, an accurate measurement of the position of the two referentials is necessary for the final readjustment of the inertial referential. It is possible by exploiting the measurements of the sights (determination of the orientation of the incident beams injected). This association makes it possible to constitute a set of so-called high measures frequencies (inertial measurements) corrected for its drifts (stellar measurements according to the invention).

Going back to the thumbnail addition process described more top, the calculation of the energy barycenter (measurement mode), can be preceded by the execution of a search mode on the cumulative image for center the measurement pattern, as already mentioned. This measurement mode can be conventionally performed on a reduced pattern of pixels. However, account given a priori the low S / N ratios obtained, it may be advantageous in the end, to use a selective measurement mode which consists in not carrying out the barycenter computation only on the signal pixels of stronger level and agglomerated at the pixel designated by the search mode, these pixels therefore a priori belonging to the image of the star. On this subject we can refer to European Patent No. 0 499 319 A1 in the name of the applicant and N.V. Philips' Gloeilampenfabrieken (PHF91 / 504).

The invention is not limited to the embodiments described above.

The principle of adding images previously described can also be applied to the simultaneous measurement of the position of the centers energy of luminous objects with large differences in intensity which can be greater than the dynamic range of the detector.

It allows in particular to measure the position of objects brilliants in rapid movement (vibration, noise, ...) in a repository slowly evolving over time (drift) and identified by sources weak light. The viewfinder is then used as a distance meter.

The invention finds a particularly application advantageous in the field of stellar registration of the line of sight of a space observation instrument through a star viewfinder.

In this context of stellar registration and in the presence of rapid movements of the entire supporting platform the instrument and the viewfinder, this process can advantageously be completed by inertial assistance to assist the viewfinder and to interpolate star measurements.

To reset the aiming axis of the instrument, this must be identified by one (or more) collimated beam (s) (test pattern), whose orientation is determined in relation to one (or more) star (s) of the sky.

The injection of the sights in the field of view is achievable by a beam mixer of the same type as that proposed for the previous optical copy.

With existing viewfinders, the measurement frequency is imposed by the dimly luminous object present in the field of sight, in general the star, and the intensity of the artificial target is adapted to these measurement conditions. We are thus led to measuring devices low frequencies, typically less than a few Hertz. This does not therefore does not allow the oscillations corresponding to the vibrations to be measured and deformations of satellite structures, occurring at a few tens of hertz.

One solution to this problem is also to adopt the measurement mode with image accumulation, the test pattern being continued at the frame frequency (a few tens of Hz) and the star over a period sufficiently long (of the order of a few seconds), compromise between precision sought, therefore the signal-to-noise ratio, and the amplitude of the drift to readjust.

In this case, the integration time of each image frame is adapted to the target signal. The information corresponding to this target being used on each frame, it is thus possible to obtain a measurement of satisfactory position in accuracy and frequency of the reference linked to the observation instrument. Bandwidth is then limited by detector possibilities, typically 50 Hz for a TV detector. This frequency can be higher by adopting specific modes of reading (reduction of the area used for example).

The measurement of the star allows you to readjust the average value of the movement of the target during the integration time, thus correcting the drifts of the whole. This assumes that the reference is not affected other movement than this drift. Otherwise, it is however possible to use another complementary means of measurement, for example example an inertial measurement, allowing simultaneously to measure the high frequency movements of the viewfinder to correct the average value set up, and if necessary, assist the viewfinder in its function of stellar sight.

Claims (8)

1. Star sight for analysing the image of at least one luminous object, comprising a position sensor which employs a charge-coupled detector array (CCD) of photosensitive elements and an image memory into which is transferred, after each integration time Ti and during a read-out time Tl, in the form of digital samples, the image detected by the said array and where it replaces the previous image, for the detection and image analysis of rapidly moving weakly luminous objects, the said star sight having, at any instant, available, as regards the displacement of the said luminous object with respect to the said position sensor, assistance information delivered by means for locating a reference frame tied to the said luminous object, comprising a digital memory window section for storing at least one window containing the image of the said luminous object, characterized in that the said digital memory window section is dedicated to the storing of a window, containing the image of the said luminous object, of specified dimensions and of variable position on the said detector array, in that it furthermore comprises selection means for selecting, during each integration time Ti_n+1, from the image contained in memory the said repositioned window relating to the previous integration time Ti_n, means of addition for adding, pixel sample by pixel sample, the said repositioned window to the window previously contained in the said digital memory section and for filling the latter with the addition result thus obtained, before the end of the integration of the image integrated at time Ti_n+1, the total number of windows accumulated that makes it possible to obtain a signal-to-noise ratio which is satisfactory for the image of the sighted star, being equal to P, and in that it is provided with computation means for determining the position of the image of the luminous object in the said array, by computation of barycentres and vector addition of these barycentres.
2. Star sight according to Claim 1, in which the total number P of accumulated windows is between 4 and 64.
3. Star sight according to Claim 1 or 2, for which the said reference frame locating means consist of an inertial facility external to the sight, which continually delivers attitude information directly to the sight.
4. Star sight according to Claim 1 or 2, charac-terized in that the said reference frame locating means consist of at least one specific, medium-class inertial sensor built into the said position sensor of the star sight.
5. Star sight according to Claim 1 or 2, characterized in that the said reference frame locating means, designed to determine the relative motion between a mechanical reference frame and the reference frame of the sight, consist of a reticle projector integral with the said mechanical reference frame and an optical invariant which provides for the copying of reticle(s) into the field of view of the sight, the latter comprising assistance computation means for deducing, from the position of the reticle(s) in the said detector array, the said assistance information necessary for selecting each of the P repositioned windows.
6. Method for analysing the image of at least one star, using a star sight with an array of CCD elements according to one of Claims 1 to 4, comprising an image memory, into which is transferred, after each integration time Ti, during a transfer time Tt, in the form of digital samples, the image detected by the said array, also comprising a central processing unit intended for the computations and a sequencer, characterized in that, for the detection of a rapidly moving weakly luminous star, an accumulation of P imagettes into an equal number of repositioned windows with predetermined dimensions, in the said CCD array, over the said star, is performed, which comprises the following steps:

   during the time Ti_n of integration of the image of order n the preprocessing, by the central processing unit, of the assistance information, consisting in manifesting in the reference frame of the sight the said information on the relative motion detected between the sight and the image which it receives on its CCD array,

   during the subsequent time Tt_n, shorter than Ti_n, the transfer into analogue memory of the array of the image of order n of the array integrated over the time Ti_n,

   then the assistance computation, for the determination of the position of the window that contains the imagette of order n,

   then the reading out of the image of order n during the time Tl_n which terminates before the end of the time Ti_n+1,

   for a time Ta_n, lying within the duration Tl_n, the acquisition of the imagette of order n by selection means actuated on the basis of the results of the assistance computation, and

   the summation of the imagette of order n with the previous imagettes whose sum is contained in a window memory section, and the replacing in this section of the old sum imagette, of order n-1, by the new one, of order n, as well as the storing in the said section of the position of the window of order n.
7. Method for accumulating imagettes according to Claim 6, characterized in that the said central processing unit determines the mean position of the said star over the array of CCD elements by vector addition of the geometric barycentre vector of the P window positions over the said array and of the energy barycentre vector which coincides with the position of the luminous object analysed in the sum imagette of order P.
8. Application of the star sight, mounted on a spacecraft platform, according to one of Claims 1 to 5, to the stellar repositioning of the said spacecraft by measuring the drift in its attitude with respect to a star.

Images